Semantic Attention Flow Fields for Monocular Dynamic Scene Decomposition
This work addresses the challenge of automatically separating foreground objects from background in dynamic scenes for applications in computer vision and robotics, representing an incremental advance over prior methods.
The paper tackles the problem of unsupervised decomposition of dynamic scenes from monocular video by reconstructing a neural volume with semantic and attention features, achieving competitive performance compared to supervised methods and improving foreground/background segmentation over existing static/dynamic split methods.
From video, we reconstruct a neural volume that captures time-varying color, density, scene flow, semantics, and attention information. The semantics and attention let us identify salient foreground objects separately from the background across spacetime. To mitigate low resolution semantic and attention features, we compute pyramids that trade detail with whole-image context. After optimization, we perform a saliency-aware clustering to decompose the scene. To evaluate real-world scenes, we annotate object masks in the NVIDIA Dynamic Scene and DyCheck datasets. We demonstrate that this method can decompose dynamic scenes in an unsupervised way with competitive performance to a supervised method, and that it improves foreground/background segmentation over recent static/dynamic split methods. Project Webpage: https://visual.cs.brown.edu/saff